From: Ronen Friedman <rfriedma@redhat.com>
Date: Wed, 8 Oct 2025 16:54:40 +0000 (-0500)
Subject: osd/scrub: new/modified perf counters for scrub preemption
X-Git-Url: http://git-server-git.apps.pok.os.sepia.ceph.com/?a=commitdiff_plain;h=bae81a5c12bc9cf8a6f98a10741801f86bf16204;p=ceph.git

osd/scrub: new/modified perf counters for scrub preemption

A new 'scrub_[replicated/ec]_io_intersects' counter is added, to count
the times a client write op intersects the active scrub chunk range.
A new 'scrub_[replicated/ec]_io_blocked' counter counts the times
the write op did not preempt the scrub, and instead blocked for the
scrub to release the chunk.

The existing labeled counter <label>_write_blocked is still there,
but will be removed in 'Umbrella'.  Its meaning was modified,
to mimic the new 'scrub_[replicated/ec]_io_blocked' counter.

Signed-off-by: Ronen Friedman <rfriedma@redhat.com>
---

diff --git a/src/osd/osd_perf_counters.cc b/src/osd/osd_perf_counters.cc
index 26a3a74fcab..85a4fe37fd5 100644
--- a/src/osd/osd_perf_counters.cc
+++ b/src/osd/osd_perf_counters.cc
@@ -407,6 +407,14 @@ PerfCounters *build_osd_logger(CephContext *cct) {
       l_osd_scrub_rppool_failed_elapsed,
       "failed_scrubs_replicated_elapsed",
       "time to scrub failure replicated");
+  osd_plb.add_u64_counter(
+      l_osd_scrub_rppool_write_intersects,
+      "scrub_replicated_io_intersects",
+      "client write op intersects chunk range");
+  osd_plb.add_u64_counter(
+      l_osd_scrub_rppool_write_blocked,
+      "scrub_replicated_io_blocked",
+      "write op did not preempt the scrub");
 
   // the replica reservation process - replicated pool
   osd_plb.add_u64_counter(
@@ -456,6 +464,14 @@ PerfCounters *build_osd_logger(CephContext *cct) {
   osd_plb.add_time_avg(
       l_osd_scrub_ec_failed_elapsed, "failed_scrubs_ec_elapsed",
       "time to scrub failure ec");
+  osd_plb.add_u64_counter(
+      l_osd_scrub_ec_write_intersects,
+      "scrub_ec_io_intersects",
+      "client write op intersects chunk range");
+  osd_plb.add_u64_counter(
+      l_osd_scrub_ec_write_blocked,
+      "scrub_ec_io_blocked",
+      "write op did not preempt the scrub");
 
   // the secondaries reservation process - EC
   osd_plb.add_u64_counter(
diff --git a/src/osd/osd_perf_counters.h b/src/osd/osd_perf_counters.h
index 0eae8fad8cd..4948b361499 100644
--- a/src/osd/osd_perf_counters.h
+++ b/src/osd/osd_perf_counters.h
@@ -170,6 +170,8 @@ enum osd_counter_idx_t {
   l_osd_scrub_rppool_successful_elapsed, ///< time to complete a successful scrub
   l_osd_scrub_rppool_failed, ///< failed scrubs count
   l_osd_scrub_rppool_failed_elapsed, ///< time from start to failure
+  l_osd_scrub_rppool_write_intersects, ///< client write op intersects chunk range
+  l_osd_scrub_rppool_write_blocked, ///< write op did not preempt the scrub
 
   // ----   scrub reservation process - replicated pools
 
@@ -196,6 +198,8 @@ enum osd_counter_idx_t {
   l_osd_scrub_ec_successful_elapsed, ///< time to complete a successful scrub
   l_osd_scrub_ec_failed, ///< failed scrubs count
   l_osd_scrub_ec_failed_elapsed, ///< time from start to failure
+  l_osd_scrub_ec_write_intersects, ///< client write op intersects chunk range
+  l_osd_scrub_ec_write_blocked, ///< write op did not preempt the scrub
 
   // ----   scrub reservation process - EC
 
diff --git a/src/osd/scrubber/pg_scrubber.cc b/src/osd/scrubber/pg_scrubber.cc
index eb9c0c697e3..222e80c5080 100644
--- a/src/osd/scrubber/pg_scrubber.cc
+++ b/src/osd/scrubber/pg_scrubber.cc
@@ -1020,7 +1020,8 @@ bool PgScrubber::write_blocked_by_scrub(const hobject_t& soid)
     return false;
   }
 
-  get_labeled_counters()->inc(scrbcnt_write_blocked);
+  const auto& unlabeled_cntrs_idx = get_unlabeled_counters();
+  get_osd_perf_counters()->inc(unlabeled_cntrs_idx.write_intersects);
   dout(20) << __func__ << " " << soid << " can preempt? "
 	   << preemption_data.is_preemptable() << " already preempted? "
 	   << preemption_data.was_preempted() << dendl;
@@ -1042,6 +1043,10 @@ bool PgScrubber::write_blocked_by_scrub(const hobject_t& soid)
 
     return false;
   }
+
+  get_osd_perf_counters()->inc(unlabeled_cntrs_idx.write_blocked);
+  // to be removed in version 'Umbrella':
+  get_labeled_counters()->inc(scrbcnt_write_blocked);
   return true;
 }
 
diff --git a/src/osd/scrubber/pg_scrubber.h b/src/osd/scrubber/pg_scrubber.h
index e7174c55db6..983804dfc30 100644
--- a/src/osd/scrubber/pg_scrubber.h
+++ b/src/osd/scrubber/pg_scrubber.h
@@ -150,6 +150,8 @@ static inline constexpr ScrubCounterSet io_counters_replicated{
   .successful_elapsed = l_osd_scrub_rppool_successful_elapsed,
   .failed_cnt = l_osd_scrub_rppool_failed,
   .failed_elapsed = l_osd_scrub_rppool_failed_elapsed,
+  .write_intersects = l_osd_scrub_rppool_write_intersects,
+  .write_blocked = l_osd_scrub_rppool_write_blocked,
   // replica-reservation-related:
   .rsv_successful_cnt = l_osd_scrub_rppool_reserv_success,
   .rsv_successful_elapsed = l_osd_scrub_rppool_reserv_successful_elapsed,
@@ -175,6 +177,8 @@ static inline constexpr ScrubCounterSet io_counters_ec{
   .successful_elapsed = l_osd_scrub_ec_successful_elapsed,
   .failed_cnt = l_osd_scrub_ec_failed,
   .failed_elapsed = l_osd_scrub_ec_failed_elapsed,
+  .write_intersects = l_osd_scrub_ec_write_intersects,
+  .write_blocked = l_osd_scrub_ec_write_blocked,
   // replica-reservation-related:
   .rsv_successful_cnt = l_osd_scrub_ec_reserv_success,
   .rsv_successful_elapsed = l_osd_scrub_ec_reserv_successful_elapsed,
diff --git a/src/osd/scrubber_common.h b/src/osd/scrubber_common.h
index b814d01fa85..9158c30762f 100644
--- a/src/osd/scrubber_common.h
+++ b/src/osd/scrubber_common.h
@@ -322,6 +322,8 @@ struct ScrubCounterSet {
   osd_counter_idx_t successful_elapsed; ///< time to complete a successful scrub
   osd_counter_idx_t failed_cnt; ///< failed scrubs count
   osd_counter_idx_t failed_elapsed; ///< time from start to failure
+  osd_counter_idx_t write_intersects; ///< client write op intersects chunk range
+  osd_counter_idx_t write_blocked; ///< write op did not preempt the scrub
   // reservation process related:
   osd_counter_idx_t rsv_successful_cnt; ///< completed reservation processes
   osd_counter_idx_t rsv_successful_elapsed; ///< time to all-reserved